CN104199884A - Social networking service viewpoint selection method based on R coverage rate priority - Google Patents

Abstract

The invention discloses a social networking service viewpoint selection method based on R coverage rate priority. The core concept of the method is that the R coverage rate of a viewpoint set in the network serves as the basis of judging the positioning performance of viewpoints, under the condition of an appointed number of the viewpoints, the set of nodes with the maximum R coverage rate in the network is selected as the viewpoints, and the set of viewpoints is made to reach the highest positioning accuracy with as little as calculation consumption. The viewpoint selection method is used for positioning a spread information source, and has higher positioning accuracy under the condition of the same number of viewpoints. According to the social networking service viewpoint selection method based on R coverage rate priority, an optimal viewpoint set can be found and has higher positioning accuracy under the condition of a fixed number of viewpoints. Fewer viewpoints are needed under the condition of guaranteeing positioning accuracy, and calculation consumption is less too.

Description

A kind of based on the preferential social networks observation point choosing method of R coverage rate

Technical field

The invention belongs to social networks technical field, be specially a kind of based on the preferential social networks observation point choosing method of R coverage rate.

Background technology

Be accompanied by a large amount of appearance of the novel social networking service such as blog (Blog), microblogging (Micro-Blog), social networks (Social Networks Services, SNS) has become one of important channel of society obtaining information.When the Information Communication on social networks brings convenience for people, also the diffusion for network rumour provides a kind of approach.Therefore need to position diffusion of information source in social networks, and then public sentiment is monitored.A feasible localization method, is in network, to dispose observation point, and information source is carried out to likelihood estimation.

Existing observation point choosing method is that Selection Center eigenwert is larger in network node is as observation point.The observation point obtaining by this method, its locating accuracy is low, and calculates consumption greatly, is not suitable for huge social networks.For above-mentioned situation, the present invention proposes a kind ofly based on the preferential social networks observation point choosing method of R coverage rate, its objective is the accuracy rate that improves diffuse source location.The observation point set obtaining by the method, the in the situation that of Orientation observation point quantity, its locating accuracy is higher; In the situation that guaranteeing locating accuracy, still less, calculating consumes also less the observation point needing.Locating accurately the rumour diffusion source point in social networks, is a kind of effective network public-opinion monitoring means.Existing a kind of localization method, is in network, to dispose a small amount of observation point, according to the observation the information of record introduction time and import direction into, the likelihood estimator of calculated candidate information source, and then inferential information source first.The accurate positioning of this method and calculating consume, and all the deployed position in network is relevant with observation point.

Existing observation point choosing method, be from network, to choose at random an observation point for some, another kind is preferentially to choose the large node of centrality eigenwert in network (such as number of degrees centrality, betweenness centrality, tight ness rating centrality, eigenvector centrality, cluster coefficients, K-core etc.).The observation point set that these two class methods are chosen, its locating accuracy is all lower, if need to guarantee a higher locating accuracy, just need to increase the number of observation point.But along with the increase of observation point quantity, the consumption of calculating also increases thereupon.For social networks user group in large scale like this, such calculating consumption can have a strong impact on the promptness of location.

Summary of the invention

In order to solve problems of the prior art, the invention provides a kind of based on the preferential social networks observation point choosing method of R coverage rate, find one group of observation point set of optimizing, this group observation point set can meet the in the situation that of Orientation observation point quantity, and its locating accuracy is higher; In the situation that guaranteeing locating accuracy, still less, calculating consumes also less the observation point needing.The core concept of the method, be using observe the R coverage rate of point set in network as judgement observation point positioning performance according to (having proof procedure below theoretical foundation), in the situation that specifying observation point quantity, choose a group node of R coverage rate maximum in network as observation point, make this group observation point to reach the highest locating accuracy with as far as possible little calculating consumption.Its technical scheme is:

Based on the preferential social networks observation point choosing method of R coverage rate, with m, represent population scale, G represents genetic algebra, t represents current population algebraically, G (t) represents that t is for population, and size (G (t)) represents that t is for chromosome number in population

Algorithm .R coverage rate is preferentially observed point set Algorithms of Selecting

Input: genetic algebra G, population scale m

Output: one group of observation point set that R coverage rate is preferential

Comprise the following steps:

Step 1: when t=0, initialization G (0);

Step 2: if t < is G

Step 3: calculate chromosomal fitness function value in G (t): get for fitness function, T wherein _imeet $\left\{\begin{array}{cc}\mathrm{if}& x_i=0,T_i=\mathrm{\&phi;}\\ \mathrm{if}& x_i=1,T_i=T_{x_i}\end{array}\right..$ For the gene x on chromosome _i, work as x _i=0 o'clock, T _ifor sky; Work as x _i=1 o'clock, the nodes i of take did R rank spanning tree as root, is all met | E (s, x _i) | the set of the node of≤r

Step 4: G (t) is carried out to replicate run, deposit father's chromosome in G (t+1);

Step 5: if size (G (t)) is < m;

Step 6: carry out interlace operation, deposit newly-generated chromosome in G (t+1);

Step 7: carry out mutation operation, deposit newly-generated chromosome in G (t+1);

Step 8: otherwise

Step 9:t+1, jumps to step 2;

Step 10: otherwise

Step 11: obtain the chromosome of fitness function value maximum in current population, decoding obtains corresponding observation point set.

Compared with prior art, beneficial effect of the present invention is:

The present invention, the in the situation that of known network topological structure and observation point quantity, applies this algorithm, can find the node set of one group of R coverage rate maximum in network.This algorithm be take genetic algorithm as basis, by the node mapping in network, is the gene in chromosome.Observation point choosing method, for the location, source that diffuses information, is compared with other observation point choosing methods, for identical observation point number.This method has its beneficial effect of higher locating accuracy and is in particular in following two aspects:

1., when the observation point quantity in network is specified, the observation point set obtaining by observation point Selection Strategy proposed by the invention, can reach higher locating accuracy, has improved the performance of network positions.

2. while needing to guarantee a higher locating accuracy in application (for example, locating accuracy can not be lower than 80%), the observation point quantity that observation point Selection Strategy so proposed by the invention needs is obviously less than existing method, can greatly reduce the calculating consumption in position fixing process.

Accompanying drawing explanation

Fig. 1 covers set schematic diagram;

Fig. 2 is the process of interlace operation;

Fig. 3 is the process of mutation operation.

Embodiment

Below in conjunction with the drawings and specific embodiments, technical scheme of the present invention is described further.

The present invention propose based on the preferential social networks observation point choosing method of R coverage rate, its objective is the observation point set of finding one group to optimize, this group observation point set can meet the in the situation that of Orientation observation point quantity, its locating accuracy is higher; In the situation that guaranteeing locating accuracy, still less, calculating consumes also less the observation point needing.The core concept of the method, be using observe the R coverage rate of point set in network as judgement observation point positioning performance according to (having proof procedure below theoretical foundation), in the situation that specifying observation point quantity, choose a group node of R coverage rate maximum in network as observation point.

Further, in order to obtain a group node of R coverage rate maximum in network, the present invention proposes based on the preferential observation point set Algorithms of Selecting of R coverage rate.The in the situation that of known network topological structure and observation point quantity, apply this algorithm, can find the node set of one group of R coverage rate maximum in network.This algorithm be take genetic algorithm as basis, by the node mapping in network, is the gene in chromosome, and particular content is as follows:

With m, represent population scale, G represents genetic algebra, and t represents current population algebraically, and G (t) represents that t is for population, and size (G (t)) represents that t is for chromosome number in population.

Algorithm.R coverage rate is preferentially observed point set Algorithms of Selecting

Input: genetic algebra G, population scale m

Output: one group of observation point set that R coverage rate is preferential

BEGIN

1. when t=0, initialization G (0);

2.IF?t＜G

3. calculate chromosomal fitness function value in G (t)

4. couple G (t) carries out replicate run, deposits father's chromosome in G (t+1);

5.IF?size(G(t))＜m：

6. carry out interlace operation, deposit newly-generated chromosome in G (t+1);

7. carry out mutation operation, deposit newly-generated chromosome in G (t+1);

8.ELSE

9.t+1, jumps to step 2;

10.ELSE

11. obtain the chromosome of fitness function value maximum in current population, and decoding obtains corresponding sight

Examine a set;

END

For a social networks, application R coverage rate preferentially observes that point set Algorithms of Selecting obtains, and is a group of specifying under observation point quantity and optimizes observation point set.This group observation point set is deployed in network, records each observation point and receive that first information introduction time and the information of information imports direction into, the likelihood estimator of the candidate's source point (non-observation point node) in just can computational grid candidate's source point of estimated value maximum, is the diffusion of information source point of estimation.Specific formula for calculation is as follows:

$\hat{s}=\frac{\exp (-\frac{1}{2}{(d-{\mathrm{\&mu;}}_s)}^T{\mathrm{\&Lambda;}}_s^{-1}(d-{\mathrm{\&mu;}}_s))}{\sqrt{\left|{\mathrm{\&Lambda;}}_s\right|}}$

Wherein, [d] _k=t _k+1-t ₁, [μ _s] _k=μ (| p (s _i, o _k+1) |-| p (s _i, o ₁) |) p (u, v) represents that u is to the shortest path between v, | p (u, v) | represent the length of this shortest path; μ represents that in network, information propagates into the average of another node required time, σ from a node ²represent variance.

Theoretical foundation proves:

In order to obtain a kind of effective observation point choosing method, the present invention starts with from the relation between observation point deployed position and locating accuracy, by analysis and observation, put deployed position impact is thought in customizing messages source and arbitrary information source locating accuracy, obtain a kind of based on the preferential observation point Selection Strategy of R coverage rate.Detailed process is as follows:

For network G and observation point set the accuracy rate of definition information source point location is:

Definition 1 (locating accuracy of specific source point).Making diffusion of information source point is s _i, independently carry out n time Information Communication, if the expection source point obtaining based on location algorithm think to locate and hit, remember that the number of times that in n experiment, hit location is m, claim based on observation point set O, s _ilocating accuracy be

Definition 2 (locating accuracy of source point arbitrarily).Choose at random x candidate's source point s in network _i, independently carrying out x time Information Communication, note hit-count is y, claims that the locating accuracy of network G based on observation point set O is P _o=y/x.

Owing to cannot predicting propagation source point in real network, so the present invention mainly considers the locating accuracy for any source point, and hypothesis network G temporal evolution not.Because being determined number and the deployment strategy of observation point to a great extent, locating accuracy affects, so the locating accuracy of the present invention's specific source point from research network and the relation between observation point deployed position are started with, analysis and observation point is disposed the relation with locating accuracy.

Localization method based on observation point, its theoretical foundation is based upon on shortest path hypothesis basis, when calculating likelihood estimator, suppose that information propagates along shortest path between node, and by the theoretical value of comparative information propagation delay (difference of information time of arrival of observation point record) and the actual value of observing, obtain the likelihood estimator of candidate's source point.The similarity that theoretical propagation delay and actual propagation postpone is higher, and the likelihood estimator error of calculating is lower, therefore can obtain following theorem.

For one group of observation point getting s is a certain designate candidate source point, and p (m, n) represents the shortest path between node m and n, supposes o ₁observation point for nearest apart from s, has following theorem

Theorem 1.If two different observation point set O ₁and O ₂, its locating accuracy with respect to s is respectively with so as l (s, O ₁) > l (s, O ₂) time, have

Proof:

The a certain s ∈ G in network G of take is candidate's source point, and message is at the unknown t constantly ^*start to propagate o ₁and o _irespectively at moment t ₁and t ₁receive message, because each limit propagation delay θ in network _imeet θ-N (μ, σ ²), have

$t_1=t^{*}+\underset{{\mathrm{\&theta;}}_i\&Element;p(s,o_1)}{\mathrm{\&Sigma;}}{\mathrm{\&theta;}}_i$

$t_i=t^{*}+\underset{{\mathrm{\&theta;}}_i\&Element;p(s,o_k)}{\mathrm{\&Sigma;}}{\mathrm{\&theta;}}_i$

$t_k-t_1=\underset{{\mathrm{\&theta;}}_i\&Element;p(s,o_k)}{\mathrm{\&Sigma;}}{\mathrm{\&theta;}}_i-\underset{{\mathrm{\&theta;}}_i\&Element;p(s,o_1)}{\mathrm{\&Sigma;}}{\mathrm{\&theta;}}_i$

If for the p based on O (s, o _i) and p (s, o ₁) the propagation delay θ of top _iarithmetic equal value, have

${\stackrel{\&OverBar;}{\mathrm{\&theta;}}}_o=(\underset{{\mathrm{\&theta;}}_i\&Element;p(s,o_k)}{\mathrm{\&Sigma;}}{\mathrm{\&theta;}}_i-\underset{{\mathrm{\&theta;}}_i\&Element;p(s,o_1)}{\mathrm{\&Sigma;}}{\mathrm{\&theta;}}_i)/\left(\right|p(s,o_i)|-|p(s,o_1)\left|\right)$

Character from expectation with variance

$\begin{array}{c}E\left({\stackrel{\&OverBar;}{\mathrm{\&theta;}}}_o\right)=E\left[\left(\frac{1}{\left|p(s,o_i)\right|-\left|p(s,o_1)\right|}(\underset{{\mathrm{\&theta;}}_i\&Element;p(s,o_k)}{\mathrm{\&Sigma;}}{\mathrm{\&theta;}}_i-\underset{{\mathrm{\&theta;}}_i\&Element;p(s,o_1)}{\mathrm{\&Sigma;}})\right)\right]\\ =\frac{1}{\left|p(s,o_i)\right|-\left|p(s,o_1)\right|}[\underset{{\mathrm{\&theta;}}_i\&Element;p(s,o_k)}{\mathrm{\&Sigma;}}E\left({\mathrm{\&theta;}}_i\right)-\underset{{\mathrm{\&theta;}}_i\&Element;p(s,o_1)}{\mathrm{\&Sigma;}}E\left({\mathrm{\&theta;}}_i\right)]\\ =\frac{1}{\left|p(s,o_i)\right|-\left|p(s,o_1)\right|}\left(\right|p(s,o_i)|\&CenterDot;\mathrm{\&mu;}-|p(s,o_1)|\&CenterDot;\mathrm{\&mu;})=\mathrm{\&mu;}\end{array}$

$\begin{array}{c}D\left({\stackrel{\&OverBar;}{\mathrm{\&theta;}}}_o\right)=D\left[\left(\frac{1}{\left|p(s,o_i)\right|-\left|p(s,o_1)\right|}(\underset{{\mathrm{\&theta;}}_i\&Element;p(s,o_k)}{\mathrm{\&Sigma;}}{\mathrm{\&theta;}}_i-\underset{{\mathrm{\&theta;}}_i\&Element;p(s,o_1)}{\mathrm{\&Sigma;}}{\mathrm{\&theta;}}_i)\right)\right]\\ =\frac{1}{{\left(\right|p(s,o_i)|-|p(s,o_1)\left|\right)}^2}[\underset{{\mathrm{\&theta;}}_i\&Element;p(s,o_k)}{\mathrm{\&Sigma;}}D\left({\mathrm{\&theta;}}_i\right)+\underset{{\mathrm{\&theta;}}_i\&Element;p(s,o_1)}{\mathrm{\&Sigma;}}D\left({\mathrm{\&theta;}}_i\right)]\\ =\frac{1}{{\left(\right|p(s,o_i)|-|p(s,o_1)\left|\right)}^2}\left(\right|p(s,o_i)|\&CenterDot;{\mathrm{\&sigma;}}^2+|p(s,o_1)|\&CenterDot;{\mathrm{\&sigma;}}^2)\\ =\frac{\left|p(s,o_i)\right|+\left|p(s,o_1)\right|}{{\left(\right|p(s,o_i)|-|p(s,o_1)\left|\right)}^2}{\mathrm{\&sigma;}}^2\end{array}$

Utilize Chebyshev inequality to obtain

$P\left\{\right|{\stackrel{\&OverBar;}{\mathrm{\&theta;}}}_o-\mathrm{\&mu;}|<\mathrm{\&epsiv;}\}\&GreaterEqual;1-\frac{\left(\right|p(s,o_i)|+|p(s,o_1)\left|\right){\mathrm{\&sigma;}}^2}{{\left(\right|p(s,o_i)|-|p(s,o_1)\left|\right)}^2{\mathrm{\&epsiv;}}^2}$

Wherein, ε is positive count, when | p (s, o _i)-| p (s, o ₁) | during → ∞, have therefore have

$\mathrm{lim\; P}\left\{\right|{\stackrel{\&OverBar;}{\mathrm{\&theta;}}}_o-\mathrm{\&mu;}|<\mathrm{\&epsiv;}\}=1$

Illustrate and work as | p (s, o _i)-| p (s, o ₁) | during → ∞, arithmetic equal value infinite approach mathematical expectation μ, has [d] _k≈ [μ] _k.

Therefore, as l (s, O ₁) > l (s, O ₂) time, have ? ratio closer to μ, therefore based on O ₁actual information propagation delay and the error between theoretical information propagation delay less.Because the information locating method that the present invention adopts, to postpone to realize with respect to the probability density distribution of actual information propagation delay by calculating theoretical Information Communication, therefore the error between actual information propagation delay and theoretical information propagation delay is less, and locating accuracy is higher.So, for O ₁and O ₂, have

Prove complete.

Theorem 1 shows, for a certain customizing messages source, observation point is to the range difference of this information source and when larger, theoretical propagation delay can reflect the truth in Information Communication process more accurately, the similarity of appointed information source in computation process is also higher, and the probability that is chosen as actual information source is also just larger.That is to say, also just higher for the locating accuracy of this source point.

If there is one group of observation point set to meet, for each designate candidate source point, all there is higher locating accuracy, this group observation point is disposed higher for the locating accuracy of arbitrary information source so.The conclusion of take in theorem 1, as basis, obtains theorem 2.

Theorem 2.If any candidate's source point s in network G _ito the distance apart from its nearest observation point, be for one group of observation point O, in candidate's source point set s maximal value so for two observation point set O ₁and O ₂, its corresponding locating accuracy is with , work as so time, have

Proof:

For one group of observation point O, get any two candidate's source point s in G _iand s _j, o _iand o _jrepresent apart from s respectively _iand s _jnearest observation point, has $\left|p(s_i,o_i)\right|=p_{{\min s}_i}\&le;r_o,\left|p(s_j,o_j)\right|=p_{{\min s}_j}\&le;r_o,$ So, s _i, s _jand o _jformed a triangle, the character according to triangle edges, has

$\left|p(s_i,o_j)\right|\&GreaterEqual;\left|p(s_i,s_j)\right|-p_{\min s_j}$

Wherein, work as o _jat p (s _i, s _j) when upper, therefore, work as s _iduring for candidate's source point, s _ito o ₁with s _ito o _jbetween path difference meet

$\left|p(s_i,o_j)\right|-\left|p(s_i,o_i)\right|\&GreaterEqual;\left|p(s_i,s_j)\right|-p_{{\min s}_j}-p_{{\min s}_i}$

$\begin{array}{c}l(s_i,O)=\underset{j=1,j\&NotEqual;i}{\overset{K}{\mathrm{\&Sigma;}}}\left(\right|p(s,o_j)|-|p\left({s,o}_1\right)\left|\right)\\ \&GreaterEqual;\left[\underset{j=1,j\&NotEqual;i}{\overset{K}{\mathrm{\&Sigma;}}}\right|p(s_i,s_j)|-\underset{j=1,j\&NotEqual;i}{\overset{K}{\mathrm{\&Sigma;}}}{p}_{{\min s}_j}-(K-1)p_{{\min s}_i}]\end{array}$

The average path length of getting nodes is R, because so

l(s _i，O)≥(K-1)(R-2r)

So, for two observation point set O ₁and O ₂, when time, there is l (s _i, O ₁) < l (s _i, O ₂), by theorem 1, can be drawn, as l (s _i, O ₁) < l (s _i, O ₂) time, have that is to say, for a certain appointed information source s _i, when time, for O ₁and O ₂, locating accuracy and for each candidate's source point s _i, all have $P_{O_1{s}_i}<P_{O_2{s}_i},$ So $P_{O_1}<P_{O_2}.$

Prove complete.

Theorem 2 shows, for one group of observation point set, if for each candidate's source point, less apart from the distance between its nearest observation point and this node, and locating accuracy of this group observation point is higher so.If one group of observation point can meet one of candidate's source point arbitrarily more among a small circle in, all there is at least one observation point, this group observation point is disposed and is one group of optimization and disposes so.

By theorem 2, can be drawn, for one group of observation point, the candidate source point less apart from observation point distance is more, and the locating accuracy of this group observation point is higher so.That is to say, observation point for specified quantity, if take a distance to a declared goal as radius (this distance is as much as possible little), the point of take in observation point set is done several circles and is removed the candidate's source point in coverage diagram as the center of circle, can cover so one group of observation point that candidate's source point is maximum is one group of observation point that locating accuracy is the highest, and the observation point that is one group of optimum is disposed.In order to obtain optimum observation point, dispose, the present invention proposes, by calculating the R coverage rate of one group of observation point set, to weigh the locating accuracy of this group observation point.The R coverage rate of observation point set is defined as follows:

Definition 3[R coverage rate].In network G, for a certain observation point o _i, all satisfied | E (s, o _i) | the set of the node of≤r be called observation point o _ir cover set.Set be called the covering set of observation point set O, claim r coverage rate for observation point set O.

As shown in Figure 1,1 coverage rate of one group of observation point of take is example, in network, choose observation point set for 1,2,5,14}, meets | E (s, o _i) | candidate's source point set of≤1 1,2,3,5,8,11,14,16,17,18,19} is one 1 of this observation point set and covers set, and its 1 coverage rate is | { 1,2,3,5,8,11,14,16,17,18,19}|/20=0.55

Obviously, along with the increase of Co, can have more candidate's source point to meet within the scope of apart from its R and have at least one observation point, so for an observation point set O, along with the increase of Co, its locating accuracy Po improves.In concrete application process, the value of R will be depending on actual conditions, depend on the ratio that practical application topology of networks and observation point are shared, the fewer R value of observation point is larger, more R of observation point value is less, principle is that the covering collection of observation point can cover under the prerequisite of whole network substantially, and the value of R is the smaller the better.

Therefore, can select R coverage rate as the evaluation criterion of observation point set.For the observation point set of equal number, the set of high R coverage rate has higher locating accuracy.So, the optimization deployment issue of observation point just can be converted into the optimization problem of R coverage rate.

Algorithm steps describes in detail:

Based on above-mentioned conclusion, the present invention proposes a kind ofly based on the preferential social networks observation point choosing method of R coverage rate, for the observation point of specified quantity, choose a kind of node of R coverage rate maximum as observation point.And then, propose R coverage rate and preferentially observe point set Algorithms of Selecting, the contents are as follows:

With n dimension 0-1 vector { x ₁, x ₂..., x _nrepresent whether node in G is chosen as the state of observation point, wherein x _i=0 represents that node i is not chosen as observation point, x _i=1 represents that node i is chosen as observation point, and k is illustrated in the number of the observation point that can dispose in G, C _othe coverage rate that represents k selected observation point set O. so, the set that reaches k node of R coverage rate maximum is one group of optimization and disposes, can Prescribed Properties and objective function be

$\left\{\begin{array}{c}{\mathrm{\&Sigma;}}_{i=1}^N{x}_i\&le;k\\ x_i\&Element;\left\{\mathrm{0,1}\right\},(i=\mathrm{1,2},\&CenterDot;\&CenterDot;\&CenterDot;N)\end{array}\right.$

max?f(x ₁，x ₂，…x _N)＝maxC _o

Obviously, the problems referred to above are set covering problems, this problem has been proved to be a np complete problem. and point set is observed in the optimization that the present invention adopts genetic algorithm to choose in network. and by the node mapping in network, be the gene on chromosome, by copying, intersect, the sequence of operations such as variation, the process of simulation genetic recombination and evolution, by iteration repeatedly, until obtain final optimum results.

(1) individual coding

If chromosome length equals nodes number n, adopt scale-of-two n n dimensional vector n x _ias the genetic coding of solution space parameter, if chromosome string i position equals 1, represent that corresponding node is chosen as observation point, otherwise represent that this is not chosen as observation point. establishing population scale is m, and maximum evolutionary generation is G.

(2) fitness function

Get $f\left(x_i\right)=\left|\underset{i=1}{\overset{N}{\&cup;}}{T}_i\right|$ For fitness function, T wherein _imeet $\left\{\begin{array}{cc}\mathrm{if}& x_i=0,T_i=\mathrm{\&phi;}\\ \mathrm{if}& x_i=1,T_i=T_{x_i}\end{array}\right..$ For the gene x on chromosome _i, work as x _i=0 o'clock, T _ifor sky; Work as x _i=1 o'clock, the nodes i of take did R rank spanning tree as root, is all met | E (s, x _i) | the set of the node of≤r

(3) replicate run (select)

Calculate the summation of all chromosomal each gene respective value in population for meeting chromosome, calculate its fitness function value, and two chromosomes of functional value maximum remained into population of future generation, as father's chromosome of population of future generation.

(4) interlace operation (crossover)

As shown in Figure 2, in two father's chromosomes, retain its portion gene (retaining length chooses at random), then by the gene cross exchanged of remainder, obtain two new chromosomes and deposit the next generation in.

(5) mutation operation (mutation)

As shown in Figure 3, to father's chromosome, the corresponding value of its a certain position gene is carried out to inversion operation, then deposits the new chromosome obtaining in the next generation.

The above; it is only preferably embodiment of the present invention; protection scope of the present invention is not limited to this; anyly be familiar with those skilled in the art; in the technical scope disclosing in the present invention, the simple transformation of the technical scheme that can obtain apparently or equivalence are replaced and are all fallen within the scope of protection of the present invention.

Claims (1)

1. based on the preferential social networks observation point choosing method of R coverage rate, it is characterized in that,

With m, represent population scale, G represents genetic algebra, and t represents current population algebraically, and G (t) represents that t is for population, size (Gt)) represent that t is for chromosome number in population,

Algorithm .R coverage rate is preferentially observed point set Algorithms of Selecting

Input: genetic algebra G, population scale m

Output: one group of observation point set that R coverage rate is preferential

Comprise the following steps:

Step 1: when t=0, initialization G (0);

Step 2: if t < is G

Step 3: calculate chromosomal fitness function value in G (t): get the R coverage value of a group node for the fitness function of this group node, be designated as $f\left(x_i\right)=\left|\underset{i=1}{\overset{N}{\&cup;}}{T}_i\right|$ T wherein _imeet $\left\{\begin{array}{cc}\mathrm{if}& x_i=0,T_i=\mathrm{\&phi;}\\ \mathrm{if}& x_i=1,T_i=T_{x_i}\end{array}\right..$ For the gene x on chromosome _i, work as x _i=0 o'clock, T _ifor sky; Work as x _i=1 o'clock, the nodes i of take did R rank spanning tree as root, is all met | E (s, x _i) | the set of the node of≤r

Step 4: G (t) is carried out to replicate run, deposit father's chromosome in G (t+1);

Step 5: if size (G (t)) is < m;

Step 6: carry out interlace operation, deposit newly-generated chromosome in G (t+1);

Step 7: carry out mutation operation, deposit newly-generated chromosome in G (t+1);

Step 8: otherwise

Step 9:t+1, jumps to step 2;

Step 10: otherwise

Step 11: obtain the chromosome of fitness function value maximum in current population, decoding obtains corresponding observation point set.